testing the relationship between the functional source of ... · the functional source of...

5

Testing the Relationship Betweenthe Functional Source of Innovation

and Expected Innovation Rents

In chapter 4 I proposed that it would often be possible to predict the func-tional sources of innovation on the basis of differences in potential innovators'expectations of innovation-related rents. Now it is time to test this hypothesis.I begin by testing the hypothesis against five samples and find it supported inthese instances. Then I draw on the evidence presented to propose that gen-eral rules may underlie innovators' expectations of rent.

Five Empirical Tests

In order to test the hypothesis that the functional source of innovation andinnovators' expectations of rent are related, we need data on both thesefactors. Since the work described in chapters 2 and 3 has already providedreliable innovation source data for nine quite diverse innovation types, Ifound it efficient to create the data base for this test by adding rent expecta-tion data to these same innovation samples where possible. After investiga-tion I found that I could obtain these additional data for five of the ninesamples (listed in Table 5-1). These became the basis of the five empiricaltests of the hypothesis I have carried out to date.

The inability to use four innovation samples due to data problems is unfortu-nate. However, the reasons for exclusion do not appear to bias our test of thehypothesis. In the instance of semiconductor process innovations, a singleinnovation caused multiple product and process impacts whose value I couldnot isolate and estimate. In the instance of plastics additive innovations andthe sample of wire termination equipment innovations, needed data weresimply not available from innovating firms or from published sources. Finally,the scientific instrument innovation sample could not be used here because

57

From The Sources of Innovation by Eric von Hippel - Oxford University Press, 1988Download courtesy of OUP at http://web.mit.edu/evhippel/www

The Sources of Innovation

TABLE 5-1. Summary of Functional Source of Innovation Data

Innovations Developed by

NA TotalInnovation Category User Manufacturer Supplier Other (n) (n)

Pultrusion process 90% 10% 0% 0% 0 10Tractor shovel-related 6 94 0 0 0 11Engineering plastics 10 90 0 0 0 5Industrial gas-using 42 17 33 8 0 12Thermoplastics-using 43 14 36 7 0 14

the innovation-related benefits expected by instrument users and instrumentmanufacturers were different in kind and could not be readily compared.(This is a problem of general interest and I will return to it at the end of thechapter.)

My hypothesis requires that I estimate the rents firms could reasonablyexpect if they had decided to develop specified innovations. This is not aneasy task because expectations of rent are not based on some straightforwardcalculation. Rather, those who plan to innovate know they must struggle forgain against the sometimes unpredictable actions of competitors, the possibleemergence of competing innovations, and other events.

In the battle for innovation-related rents, each firm with an interest in aninnovation devises strategies that may help it to minimize innovation costs andmaximize its returns from innovation. For example, if Boeing decides todevelop a new, more fuel-efficient plane, it will try to lower its innovationcosts by shifting some project development expenses to component suppliersand by demanding some advance. payments from buyers. Also, it will try toincrease its share of the rents generated by the plane by, for example, raisingits price to capture some of the fuel savings benefit that users expect to reap.At the same time, of course, suppliers and users are trying to resist thesemoves and to carry out profit-increasing strategies of their own (e.g., GeneralElectric may attempt to charge more for the fuel-efficient jet engines that itsupplies to Boeing).

Given this complex reality, my general strategy for estimating rents thatfirms might reasonably expect if they were to develop specific types of innova-tions has been to study several innovation categories in detail and to try tounderstand the thinking of and options open to potential innovators in thesefields.

Innovators capture temporary rents from their successful innovations byfirst establishing some type of monopoly control over their innovation andthen using this control to increase their economic return. A successful innova-tor's rents may come in the form of cost savings and/or increased prices and/or increased sales obtainable during his period of temporary and partialmonopoly. Unfortunately, the available data do not allow us to assign values

58

Link Between Innovation and Expected Rents

to these individual components of expected rents or to properly sum anddiscount them with respect to time. Therefore, in each test that follows I willfirst summarize information bearing on firms' relative abilities to captureinnovation-related rents and then will build a test of reason from this data.

The logic behind my tests of reason will be self-evident, I think, with thepossible exception of some of the elements I use to estimate relative abilitiesto establish monopoly control over an innovation. Let me therefore elaboratea little on the latter before proceeding to a discussion of the tests themselves.

Recall from chapter 4 that we found that innovators typically could notexpect to obtain rents from their innovations by licensing them to othersbecause both the patent grant and trade secrecy legislation did not typicallyallow innovators the type of monopoly control necessary to achieve this.Therefore, innovators must typically benefit by excluding imitators from ob-taining rent from their own innovation-related outputs.

The patent grant, trade secret legislation, and response time are the onlymechanisms I have observed to date that are exclusively available to innova-tors and may potentially give an innovator the control over his innovation thathe needs to exclude would-be imitators. (This seems a short list, but note thatit is not of fixed length. Mechanisms for enhancing an innovator's innovationproperty rights are social inventions and their number or design has no inher-ent limit. For example, in the United States one can currently observe theextension of copyright protection to include software writings.)

In chapter 4 we saw that the patent grant generally offers only weak protec-tion to innovators in the context of licensing. I find no apparent reason why itshould be any more effective in allowing an innovator to prevent others fromimitating his innovation without permission. However, trade secret protectioncan be much more effective in preventing imitation than it was seen to be inenabling licensing. The difference is that, as a practical matter, licensing oftrade secrets requires that they be revealed to licensees, and a secret sharedmay not stay secret very long. In contrast, a secret kept by an innovator for hisown exclusive use need only be known within his factory walls and can oftenbe well protected there.

Finally, an interesting and often effective additional mechanism-responsetime-became visible in the course of my investigations of innovators' strate-gies for protecting their innovation-related knowledge. I define response timeas the period an imitator requires to bring an imitative product to market or tobring an imitative process to commercial usefulness once he has full and freeaccess to any germane trade secrets or patented knowledge in the possessionof the innovator.

Response time exists because many barriers in addition to lack of knowl-edge must be overcome in order to bring any product or process-even animitative one-to commercial reality. Engineering tooling must be designed,materials and components ordered, manufacturing plants made ready, market-ing plans developed, and so on. During the response-time period an innovatorby definition has a monopoly with respect to the innovation and is in a posi-tion to capture rent from his innovation-related knowledge.

59


When an innovator is seeking to protect his innovation from imitators, hemay be able to use any or all three mechanisms-patents, trade secrecy, andresponse time-to prolong his period of exclusive use, his lead time. Whetherany or all of these are usable or offer an advantage to a given class of potentialinnovator is a function of situation-specific factors that will be discussed in thecontext of the cases that follow.

I begin each case discussion by identifying the types of firm best positionedto capture rents from the category of innovations under study. Next I describeand compare four rent-related elements of the real-world situation facing eachtype of firm. These are (1) the relative abilities of firms holding differentfunctional relationships to an innovation to establish some monopoly controlover it; (2) the nature and amount of innovation-related output generated byinnovating and noninnovating firms; (3) the anticipatable cost of innovation;(4) the displacement of existing business that a firm undertaking the innova-tion studied might expect. Finally, I draw on this information to assess thereasonable preinnovation rent expectations of potential innovators. In thisfinal step I use a very conservative test with respect to my hypothesis. I simplyrank such rent expectations and count the hypothesis as supported only if thefunctional source of innovation is populated by firms with the highest pre-innovation expectations of rent. (I will return to this point at the end of thechapter.)

Pultrusion Process Machinery:Innovation and Innovation Rents

As we saw in chapter 3, users are the source of all sampled pultrusion processmachinery innovations save one. Therefore, the hypothesis I am testing maybe stated for this sample as: Innovating users of pultrusion process machineryhad higher preinnovation expectations of rent from their innovations than didall firms holding other functional relationships to those same innovations.

There are many classes of firm that have some sort of functional relation-ship to pultrusion process machinery innovations ranging from inventor touser to manufacturer to distributor. However, it is not necessary to preciselydetermine the reasonable rent expectations of all of these in order to test thehypothesis. The only noninnovators who can represent a challenge to thehypothesis are those with rent expectations that might conceivably equal orexceed those of the innovating firms. It is therefore efficient to begin this testwith a simple inspection of the many extant functional relationships betweeninnovator and innovation and exclude all from further analysis that on theface of it cannot reasonably expect a level of rent from pultrusion processmachinery innovations that is anywhere near that of the innovating processusers.

In the instance of pultrusion process machinery innovations, I have deter-mined, by means of interviews with industry experts, that the only two func-tional categories of firm likely to gain significant rent from innovations that

60

Link Between Innovation and Expected Rents 61

improve pultrusion process machinery are process machine users and processmachine manufacturers.* I therefore explore the relative ability of these twofunctional types of firms to capture such rent in what follows.

Relative Ability to Establish Monopoly Control

On the basis of both interview data and reason, it appears to me that innovat-ing users of pultrusion process equipment are better able than innovatingmanufacturers to establish temporary monopoly control over their innova-tions. The key source of this difference is the ability of equipment users tohide their innovations for a period of time as trade secrets. This option is notopen to manufacturers, who must display their innovations to customers inorder to sell them.

Patents and response time were both considered ineffective in this industryby interviewees. Innovators sometimes applied for patents and sometimesreceived some modest licensing income, but it was understood by all that thepatents did not really provide effective protection. Innovators typically wouldnot even attempt to contest infringement of their patents and expected onlythe naive or exceedingly cautious to honor them. Response time in the in-stance of pultrusion innovations is only months and is not considered to be ofsignificant value to either user or manufacturer innovators.

Relative Innovation-Related Output

The innovation-related output of the users of pultrusion process equipmentinnovations is, of course, pultruded product. In general, the effect of theinnovations we have studied was to make it possible to extend the pultrusionprocess to new types or sizes of product. This in turn allowed user firms tocreate cheaper or better substitutes for products made by other methods orother materials.

*Field investigation showed that the licensing of innovation-related knowledge was (and wasconsidered to be by industry participants) of minor importance in pultrusion processing. Innova-tors' general inability to appropriate rents from the licensing of knowledge led me to eliminateboth independent inventors and suppliers of materials used in pultrusion as potential recipients ofsignificant innovation-related benefit. Independent inventors were eliminated because they onlyhave nonembodied innovation knowledge to sell. Suppliers of materials used in pultrusion wereeliminated because pultrusion used 2% or less of the huge amount of polyester resin and fiber-glass consumed annually in the manufacture of fiberglass-reinforced plastics during the period ofthe innovations studied (William G. Lionetta, Jr., "Sources of Innovation Within the PultrusionIndustry" [SM thesis, Sloan School of Management, MIT, Cambridge, Mass., 1977], Table III, p.41). These innovations needed only commodity plastic resin and fiberglass reinforcement prod-ucts to implement. Therefore, the suppliers too could not embody innovation-related knowledgein their outputs, and would also be dependent on the licensing of nonembodied innovationknowledge for their innovation-related rents.

Other functional relationships between innovator and pultrusion process innovations-wholesaler and so forth-were eliminated from further consideration after discussion with indus-try personnel showed innovation benefit potentially accruing to these was clearly much less thanthat potentially accruing to users and manufacturers.


Pultrusion process machines vary in size, and early machines tended to besmaller and slower than later ones. In 1975 the output of a single averagepultrusion machine working a single shift was about 200,000 lb of pultrusionsannually. At the 1976 market price of $1.70/lb, this means that a single innova-tive pultrusion machine could produce $340,000 of novel pultruded productduring each year of that machine's service life.

The innovation-related output of a pultrusion machine manufacturer is theincremental hardware on the pultruder that embodies a pultrusion improve-ment innovation. However, it is more conservative with respect to the hy-pothesis we are testing to regard innovation-related output as the entire ma-chine embodying such an innovation. Early pultrusion machines tended to besmaller and cheaper than later ones. In 1975 the price of a commercialpultruder of average capacity was $75,000 to $85,000.2

Until 1966, all pultrusion process equipment was made by the firms thatused it. Then, in 1966 Goldsworthy Engineering began to manufacture astandard line of pultrusion process machines. (Of course, prior to that yearthere were many manufacturing firms that did produce other types of plasticsprocessing machines and were fully capable of producing pultruders.) By the1970s each of the largest three user firms had 15 to 20 pultrusion machineseach and were making machines for internal use at roughly the same rate asGoldsworthy was making them for the external market. In 1976, pultruderoperators reported that about 30 of the 175 pultruders operating in the UnitedStates had been built by an equipment builder. The rest were built by users in-house. In that year, pultruders built by users in-house cost less than thecommercial equivalent.3

Relative Costs of Innovating

The pultrusion process machinery innovations examined were built by bothusers and manufacturers using general-purpose machine shop equipment thatboth types of firm had on-site. No organized R & D effort was used to developthese innovations: They consisted of good ideas that, once grasped, could beimplemented on the shop floor. Thus it can be assumed that innovation costswould be similar in magnitude for both users and manufacturers of thesemachines.

Relative Amount of Displaced Sales

No potential innovator in this field had reason to anticipate that pultrusionprocess innovations would result in a significant displacement of his presentsales. In some instances, as can be seen in the innovation cases presented inthe appendix, users developed process improvements to reduce their costs ofproduction. In other cases, pultrusions were displacing metals, typically, inhigh-performance applications. Neither the pultrusion process equipment us-ers nor the equipment manufacturers had any position I am aware of in suchdisplaced products or processes.

62


Discussion

Now I must speculate. Is it reasonable that even an optimally situated manu-facturer of pultrusion machines could expect to gain as much or more rent asinnovating users? I do not think so. We have seen that a user has a basis toexpect some degree of monopoly control over his innovation, but that amanufacturer does not. Further, I see no relative disadvantage that an innovat-ing user might have with respect to an innovating manufacturer that mightoffset the user advantage in monopoly control. Users and manufacturers con-templating innovation could expect similar costs, and neither needed to feardisplacement of existing business as a result of the type of innovation studied.Further, the possible impact of increasing returns to scale in production of themachines embodying the innovation would be trivial at the levels of produc-tion involved here. Even if a single manufacturer produced all of the machinesneeded by the market (say, 20 per year) his direct per machine costs wouldonly be on the order of 10% less than those of a user producing only a singlemachine. 4

In sum, then, the evidence leads me to conclude that the rents that processmachine users could reasonably expect prior to an innovation exceed thosethat a pultrusion process machine manufacturer could expect if he contem-plated the same innovation opportunity. Therefore the hypothesis is sup-ported in this instance: Pultrusion process users, the functional type of firmthat I judge to have the highest reasonable expectations of innovation-relatedrent, are also the type of firm that my data show most active in developingpultrusion process equipment innovations.

The Tractor Shovel: Innovation and Innovation Rents

Since, as we saw in chapter 3, manufacturers are the source of almost alltractor shovel innovations, the hypothesis I am testing may be stated for thissample as: Firms that hold the functional relationship of manufacturer to thesampled tractor shovel innovations are also the type of firm best positioned tocapture rents from such innovations.

As in the previous study, my first step in testing the hypothesis here was toidentify by inspection the few functional types of potential tractor shovelinnovators positioned to appropriate significant innovation rents. I foundtractor shovel users and tractor shovel manufacturers were clearly the twofunctional categories of firm most favorably positioned in this regard,* and Iwill therefore only attempt to rank the relative rent expectations of these twotypes of firms here.

*My findings here precisely parallel those presented earlier regarding the pultrusion processmachinery study. As was the case in that field, innovation benefit was almost never captured fromthe licensing of nonembodied knowledge. Further, industry participants had no illusions that thiscould be done, given the general weakness and unenforceability of patents in the field of mechani-cal invention. As a consequence, I judged that independent inventors were unlikely to be able to



Tractor users and tractor shovel manufacturers apparently are in an equallypoor position to establish monopoly control over a tractor shovel innovationthey may undertake. Patents did not offer effective protection to innovators inthis field. Also, neither user nor manufacturer could expect to protect tractorshovel product innovations as trade secrets. Tractor shovels are used on openconstruction sites, and any innovations by users and/or manufacturers will beopen to the view of would-be imitators. Response times in the instance oftractor shovel innovations are on the order of one year. That is, either a useror a manufacturer of a tractor shovel who had good mechanical skills couldimitate a tractor shovel innovation he was able to inspect in a year or less.


The innovation-related output of tractor shovel manufacturers consists ofhardware embodying such innovations. The advantage of the sampled majorimprovement innovations over previous best practice was such that they wereimmediately embodied in most or all of the units sold by the innovating firm.Therefore, total tractor shovel sales of the innovating firm in the first post-innovation year is a reasonable indicator of the units of output embodying theinnovation in that year.5

In the instance of the development of the tractor shovel itself, first-yearinnovation-related sales were about $250,000 (50 units were sold). All im-provement innovations were developed by manufacturers producing at leasthundreds of tractor shovels in the year the innovation was commercialized.The major improvement innovations in the sample all added functional capa-bilities to the tractor shovel at some increase in complexity and cost.

The function of a tractor shovel is to excavate and/or move bulk materials.From a user's point of view, the innovation-related output of a tractor shovelinnovation is the increase in productivity that the innovation provides. I esti-mate the improvement that an innovating user could expect from embodyingone of the improvement innovations studied on one tractor shovel is an in-crease in output of on the order of 20%. This translates into an operatingsavings of perhaps $1000 annually per machine. 6

Even the largest users of tractor shovels contemplating an innovation couldnot expect to incorporate their innovation on more than a very few tractorshovels. Today, with the exception of the U.S. Army and some municipali-ties, even the major, national account users have a fleet of only 8 to 10 tractor

appropriate significant innovation benefit from any tractor shovel innovations they might at-tempt. The same reasoning suggests that suppliers of components used to implement the innova-tions were also unlikely to innovate. Components used to achieve an innovative effect whenapplied to tractor shovels were not themselves novel and were typically available from a numberof suppliers as off-the-shelf items.

Industry experts contacted all judged that users and manufacturers ranked highest in the list offuctional types of firms able to appropriate innovation benefit from tractor shovel innovations.

64


shovels-and that of mixed models and vintages. Presumably fleets of this sizewere even more unusual in the period when the innovations we studied werecommercialized.


I estimate that user innovation costs would be somewhat higher than thoseof innovating manufacturers because the equipment and engineering skillsneeded for innovation are utilized by tractor shovel manufacturers in thecourse of routine manufacturing and are therefore in place if needed forinnovation. In contrast, tractor shovel users have no routine need for suchequipment or related engineering skills and may need to acquire them espe-cially for innovation-related tasks.


In the instance of the tractor shovel, users might find that the development of animprovement slightly reduced the value of older tractor shovels and otherfunctionally similar construction equipment that they had in inventory. In con-trast, with the exception of Clark Equipment Company (a tractor shovel manu-facturer that developed one of the sampled innovations), none of the innovat-ing equipment manufacturers also made construction equipment of similarfunction, such as bulldozers. Therefore, these manufacturers would anticipateno displaced sales as a result of developing tractor shovel innovations.

Discussion

On the basis of the above discussion, I reason that both users and manufactur-ers of tractor shovels could only expect to have a monopoly of a year's dura-tion if they chose to develop a tractor shovel innovation. However, it alsoseems clear that a tractor shovel manufacturer's ability to capture rent on thebasis of this period of temporary monopoly is greater than that of any user.

Tractor shovel manufacturers inform me that they generally tend to try toincrease sales on the basis of an innovation rather than to increase prices.They feel that they may gain a significant advantage over competitors in ayear, possibly selling hundreds of additional tractor shovels due to the pres-ence of the innovation. And, even after imitators enter, innovators' first-to-market reputation may continue to give them a marketplace advantage. 7

In contrast, it is not clear to me or to users how a year's monopoly on atractor shovel innovation might significantly benefit innovating users beyondthe small operating savings discussed earlier. Because there are many substi-tute ways to move materials, a given user's unique possession of an innovationwould not seem to give him monopoly control over any unique capabilitiesthat would possibly command a high rent from the market.

In this industry, users appeared to have no plausible offsetting advantagesthat might raise their expectations of rent relative to manufacturers. Indeed,


significant returns to scale obtainable from hardware embodying the innova-tions studied in the tractor shovel market were available to manufacturers butnot users. 8

In sum, then, I conclude that tractor shovel manufacturers are both thefunctional type of firm likely to have the highest preinnovation expectationsof innovation-related rent and also the functional type of firm that in fact didinnovate. Thus, I find support for my hypothesis in this sample.

Engineering Plastics: Innovation and Innovation Rents

Since, as we saw in chapter 3, the manufacturers of engineering plastics arethe source of almost all innovations sampled in that field, my hypothesis maybe stated here as: Manufacturers of the sampled engineering plastics innova-tions are also the functional type of firm best positioned to capture rents fromsuch innovations.

On the basis of inspection and discussion with industry experts, I concludedthat firms with either a user or a manufacturer relationship to engineeringplastic resin innovations are most likely to gain significant rents from them.*Therefore, I focus on assessing the relative rents appropriable by each ofthese two functional groups in this test.


In the instance of engineering plastics, all the innovators studied could and didprotect their innovations effectively through patents. This observation fits thegeneral evidence regarding the high effectiveness of patents in the field ofchemical inventions9 and discussion with industry personnel shows that suchprotection is an important part of the commercial strategy of innovators in thefield.

Recall from our earlier discussion in chapter 4 that any innovator couldexpect similar amounts of rent from a given innovation, given that he hadperfect, costlessly enforceable property rights to it. To the extent that thepatent protection available in the engineering plastics field has these character-istics, the reasoning applies here and I would expect user and manufacturerinnovators to have roughly similar expectations regarding the rents they mightobtain from engineering plastics innovations.

However, as was also discussed earlier, in the real world even strong pat-ents do not provide protection that is either perfect or costlessly enforced.

*Although patents offer effective protection in this field, licensing of engineering plasticsinnovations is unlikely for reasons to be discussed in the text. As a consequence, independentinventors and others without significant innovation-related outputs did not seem to me to bepotential appropriators of significant innovation-related rents. Suppliers were eliminated fromconsideration as potential innovators because the materials used in the manufacture of the plas-tics studied were commodity chemicals, whose suppliers did not appear well positioned to obtainbenefit from innovation-related knowledge through sale of a commodity output.

66


Indeed, an innovator cannot realistically expect to license his patent rights toothers without risk or cost.


The innovation-related output of an engineering plastics manufacturer is thenovel plastic itself. Manufacture of each of the engineering plastics studied ishighly concentrated. In my sample the innovator retained a dominant share ofthe market for his novel product for at least several years. In 1976 Du Pont(the innovator) produced 100% of acetal homopolymer (Delrin), Celanese100% of acetal copolymer (Celcon), General Electric 100% of modifiedpolyphenylene oxide (Noryl), and Union Carbide 100% of polysulfone. Gen-eral Electric also produced 75% of all polycarbonate (Lexan) in 1976 and DuPont produced 57% of all polyamides and 57% of all fluoropolymers.l 0

The innovation-related output of the user of engineering plastics is theproducts in which these plastics are embodied. I have no data on the volumeof engineering plastics consumed by individual users. However, I can reportthat there are literally thousands of users of each of the innovative engineer-ing plastics studied and that no single user buys more than a small fraction oftotal production. (The sole exception I identified was in the early days ofLexan production when GE used as much as half of its pilot-plant productioninternally. This fraction quickly dropped to just a few percent when largerplants were brought on line.)


The costs for the actual innovating firms in this field were either equal to orlower than those that other would-be innovators could reasonably anticipate.All of the innovating firms had substantial ongoing research programs inorganic chemistry and thus did not have the significant R & D start-up costsfirms without such programs could expect. Some, but not many, users alsohad such programs in place.

The R & D investment that innovators expend to develop an engineeringplastic is orders of magnitude higher than the R & D costs associated withother categories of innovation examined in this chapter. Du Pont's R & D andpilot plant expenditures for Delrin, for example, were $27 million in 1959dollars.'l Commercialization is also expensive, because engineering plasticsmanufacture requires special-purpose plants. Thus, the cost of the first com-mercial plant for Delrin was $15 million' 2; for Celcon $15 to $20 million' 3; andfor Lexan $11 million. 14


Engineering plastics are intended to be substitutes for other engineeringplastics or more traditional engineering materials such as metal and glass.

67


Potential innovators who currently produce such materials do risk displace-ment of sales of existing products if they innovate. However, the problemcan be minimized or eliminated by proper selection of the properties of theinnovative material so that it is not a substitute for the innovator's existingproducts.

Discussion

As mentioned above, patent protection in this field is good, but there is noreason to expect licensing to be either costless or risk-free for an innovatorattempting it. As a consequence, I reason that potential innovators wouldexpect somewhat higher rates of rent for innovation-related output producedby the innovator's own firm. Taken by itself, this fact would not create ahigher rent expectation for either user or manufacturer innovators. However,important economies of scale issues tip the balance toward the manufacturer.

Major engineering plastics such as those I studied are manufactured insingle-purpose, continuous-flow plants that have significant economies ofscale associated with them. As an approximation, processing costs per poundin a plant sized to produce 1 million lb of plastic a year are 10 times those in aplant sized to produce 100 million lb annually.15 Consider the impact of thisfactor on a user contemplating developing a new engineering plastic. Sinceany individual user represents only a small portion of the total demand for anengineering plastic, a user firm considering innovation could not expect toachieve attainable economies of scale by producing for in-house use only. If auser, nonetheless, built a plant sized to fill in-house demand only, it would bein a risky position: Any manufacturer that came up with a functional substi-tute for the user innovation and produced it in a plant sized to serve the entiremarket could render the user plant and material uneconomic.

In effect, therefore, both users and manufacturers in this industry are bothin a position to control the rights to an innovative engineering plastic that theymay develop. But only a manufacturer (or a user who becomes a manufac-turer) is in a position to exploit the significant economies of scale associatedwith engineering plastics manufacture. Given that this is so, a manufacturerthat innovates is the only functional type of firm that does not have to incurthe cost and risk of licensing this type of innovation to a manufacturer. Theconsequent saving in licensing-related cost and risk results in a higher expecta-tion of net innovation-related rent for an innovating manufacturer than thatwhich an innovating user might expect.

Process Equipment Utilizing Industrial Gasesand Thermoplastics: Innovation and Innovation Rents

The final two tests of the hypothesis were conducted by VanderWerf. 16 Thesestudies were designed precisely to test the hypothesis that rent expectations

68


and the functional sources of innovation are linked and therefore they do notrequire extra data or analyses to serve my purposes. As a consequence, thereview I will present can be relatively brief.

Recall from chapter 3 that by the time VanderWerf began his research, heand I had the hypothesis reported on here in mind. It seemed on the basis ofthat hypothesis that materials suppliers might be found to innovate under theright conditions. As a consequence, VanderWerf elected to focus on catego-ries of process machinery innovations that used large amounts of material asinput and where an innovating supplier might hope to have some level ofmonopoly control over that input material. Under such circumstances it ap-peared possible that materials suppliers might benefit from and develop inno-vative process equipment that used their material-even if they did not planto build or use the innovative equipment themselves. Thus, in the instance ofthese two studies, we were attempting to predict the functional source ofinnovation on the basis of assumptions regarding the likely functional sourceof highest innovation rents. As the reader will see, this experiment workedwell.

The Studies

VanderWerf's first sample consisted of innovations in process machines thatutilized large amounts of industrial gases as an input, whereas his secondsample consisted of innovations in process machines that used large amountsof specific thermoplastics as an input. He began his research by determining,by means of discussions with industry experts, that process machine users,manufacturers, and suppliers of materials processed on the innovative ma-chines all had some reasonable expectations of innovation-related rents.Next, he compared the levels of such benefit that these three functionalcategories of firm might reasonably expect.

VanderWerf estimated the benefit firms could potentially appropriate fromeach innovation under study by studying the actual commercial history andinnovation-related behavior of users, manufacturers, suppliers, and others.By means of discussions with industry participants, he then estimated therelative appropriable benefits (rent), innovation costs, and the new businessfraction (a measure that serves the same function as my displaced sales) eachclass of would-be innovators could reasonably expect if they had been able toaccurately foresee the commercial results actually attained by the variousinnovations. Possible error in these estimates was compensated for by resolv-ing ambiguity in a direction against the hypothesis under test.

Discussion

For each innovation, VanderWerf ranked four functional categories of innova-tor (user, manufacturer, supplier, and other) in the order of their expected

69


TABLE 5-2. Test of Hypothesis; The Source of Innovation Benefit andInnovation Activity Compareda

(A) Industrial gas-using innovations:Predicted probability of innovation

Highest Second Third Lowest

Innovator 8 4 0 0Noninnovator 4 8 12 12

(chi 2p < .01)

(B) Thermoplastics-using innovations:Predicted probability of innovation

Highest Second Third Lowest

Innovator 12 1 1 0Noninnovator 2 13 13 14

(chi2 p < .01)

aVanderWerf, "Explaining the Occurrence of Industrial Process Innovation by Materials Suppliers," 65-66.

level of benefit from that innovation. In the top (bottom) rows of Table 5-2(A) and 5-2(B), VanderWerf positions each firm that actually did (did not)develop each innovation in that expected benefit ranking. As can be clearlyseen in Table 5-2, these two samples also strongly support the hypothesisunder test.

Conclusions and Discussion

The hypothesis I set out to test was that the functional sources of innovationcould be predicted on the basis of potential innovators' expectations ofinnovation-related rents. We now see that this hypothesis is supported in theinstance of the five samples examined (Table 5-3). These test data are encour-aging but clearly cannot prove the matter beyond dispute. Nevertheless, Imyself find the results encouraging enough to warrant moving ahead to bothfurther research and practical applications.

It would be interesting, for example, to use expectations of rents to predictand empirically explore functional sources of innovation in addition to theuser, manufacturer, and supplier sources documented in this book. Thus,wholesale or retail distributors of innovative products, processes, or servicesshould have high expectations of innovation-related rents under some condi-tions, and they probably will be found to innovate where these pertain.

I speculate that the model will be applicable to industrial products, pro-cesses, and services-a very considerable universe. It will also be quite practi-cal in these fields: It requires data on variables that innovators and policymak-ers in firms may already have at hand.

70


TABLE 5-3. I There a Relationship Between the Functional Source of Innovationand Reasonable Expectations of Innovation Rent?

Number of Innovations found insources with expected rent rank

Innovation Type Relationship Highest 2nd 3rd 4th

Pultrusion process Yes: p < .02a 9 1Tractor shovel-related Yes: p < .02a 13 3Engineering plastics Yes: p < .2b (NS) 4 1Industrial gas-using Yes: p < .01c 8 4 0 0Thermoplastics-using Yes: p < .01c 12 1 1 0

aChi2 test. The null hypothesis used in these tests was that innovations would be found equally distributedbetween the two loci of highest expected rents.bBinomial test (used due to very small sample size). The innovation coded 50% user and 50% manufacturer inTable 3-6 is coded conservatively with respect to the hypothesis here (i.e., coded as 100% not in the locus ofhighest expected rents).CChi 2 test. Same as note (a) except that VanderWerf was able to rank expected rents reasonably anticipatableby members of four functional loci with respect to each innovation (see Table 5-2).

However, the model will not be practicable in fields where all potentialinnovators do not measure the rents they expect in commensurable ways. Thisis so simply because the hypothesis that underlies the model requires that onecompare levels of expected rents across functional groups in order to predictthe functional source of innovation. And this is only possible in fields whereall classes of potential innovators with a significant potential interest in aninnovation use commensurable measures. Two examples will illustrate theproblem.

In scientific instruments, innovating scientist-users typically work innonprofit institutions. Usually, their research is supported by governmentalagencies who distribute grants on the basis of the expected scientific value ofthe proposed research rather than on its expected economic value. Therefore,neither users of scientific instruments nor their employers appear to have anyreason to measure expected innovation-related benefit in economic terms.Rather, acknowledgement by peers of scientific accomplishment appears tobe a major innovation-related incentive for this group.' 7 In contrast, scientificinstrument manufacturers are profit-making firms and presumably do mea-sure their expected innovation-related benefit in economic terms. Similarly,in the field of consumer goods, consumers are known to evaluate innovationsin part in terms of psychological benefits not easily measured in economicterms. 18

In this chapter I have been able to assess the relationship between innova-tion and firms' expectations of innovation-related rents only by very carefulattention to the details of industry structure and behavior that form suchexpectations. An important next step in this work, it seems to me, is to movetoward generalization by seeking out real-world principles and common

71


strategies-successful and unsuccessful-that underlie innovators' attempts tocapture such rents.19

The empirical work I have done to date contains clues that might lead tosuch an ability to generalize. For example, note that the functional source ofinnovation was shown in the five studies to be populated by firms expectingthe highest innovation-related rents (Table 5-3). Since the hypothesis itselfproposes only that innovation will be found concentrated among firms whofind their expected rents attractive, this result is striking. Perhaps it signalsthat expectations of rent will often differ significantly between firms holdingdifferent functional relationships to a given innovation opportunity. Thisseems to me to be possible because at the moment I can see two generalreasons why the rent expectations of potential innovators could differ signifi-cantly as a consequence of the functional relationship they hold to an innova-tion opportunity.

First, the abilities of firms to protect and benefit from identical innovation-related information can differ as a consequence of functional role. For exam-ple, innovating users can often protect process and process machinery innova-tions as trade secrets better than any other type of innovator. This is sobecause only users can obtain rent from their innovations while keeping themhidden within their factory walls. Innovators with any other functional rela-tionship to an innovation such as manufacturer or supplier must sell theinnovation they develop or persuade others to adopt it before they can bene-fit. The process of selling or persuading typically involves revealing relatedsecret knowledge to prospective innovation adopters and, as a practical mat-ter, usually destroys the basis for trade secrecy protection.

Also, the risk users face in developing a cost-reducing innovation for theirown use may typically be lower than that facing any other functional categoryof innovator. This is because only users do not have to market such innova-tions in order to derive rent from them-their own use constitutes a source ofsuch rent. Therefore, the risk to users engaging in an innovation process isthat the completed device will not work as intended or will be obsoleted bysome other innovation or event. Manufacturers and all other nonusers consid-ering developing that same innovation, on the other hand, face these samerisks plus the risk that the innovation will not be accepted in the marketplace.

A second general reason why the rent expectations of potential innovatorscould differ significantly as a consequence of the functional relationship theyhold to an innovation opportunity has to do with industry competitive struc-ture. Firms having user, manufacturer, and supplier relationships to a giveninnovation often come from different industries. These industries may havedifferent structures, for example, the industry that will manufacture an innova-tion may be more concentrated than the industry that will use it. Since a firm'sinnovation-related output is not likely to be determined entirely on the basisof a particular innovation, we can see that expectations of rent are likely to beaffected by factors such as preinnovation concentration ratios.

A better general understanding of how firms capture innovation-related

72


rents would have implications beyond innovation. For example, consider thepossible impact on current views on when and why firms specialize.

Stigler20 hypothesizes that "vertical disintegration" will occur as a new indus-try grows and matures. He reasons that many functions that firms in the newindustry require have increasing returns to scale. Initially, the firms in the newindustry might perform such functions for themselves because specialization islimited by the extent of the market and because total demand for a particularfunction might not be great enough to support a specialist firm. But, as theindustry grew, demand would increase and eventually it would be reasonablefor firms in the industry to spin such functions off to specialist firms that couldcarry them out on a larger scale and thus more cheaply.

It does seem reasonable to me that Stigler's hypothesis may explain real-world behavior under circumstances in which economies of scale and consider-ations of production cost are very important. However, different patterns willemerge when innovation-related rents (a factor not included in the Stiglerhypothesis) are important-as they often are.

Consider, for example, that users may only be able to obtain rents oninnovative process machinery if they build it in-house and protect it as a tradesecret. Such rents might be far more significant than any scale-related econo-mies potentially offered by a specialist process equipment manufacturer. An-ecdotal evidence exists in support of this idea. Thus: "Most world class Ger-man and Japanese manufacturing companies have large, well-staffed, veryactive machine shops. Much of the success of these companies is a result of theproprietary production processes that are incubated in these shops and there-fore unavailable to their competitors."21

Notes

1. In principle, if an imitator became aware of an innovator's protected knowledgeat the moment he developed it, there would be no response-time protection for theinnovator: both innovator and imitator could proceed with commercialization activi-ties in tandem. Response time is an important innovation benefit capture mechanismin reality, however, because would-be imitators seldom become aware of an innova-tor's knowledge at the moment he develops it. Typically, in fact, an imitator onlybecomes aware of a promising new product when that product is introduced to themarketplace. Until that point the innovator has been able to protect his product fromthe eyes of interested competitors inside his factory. After that point, if the product iseasily reverse engineered and has no patent protection, only the response-time mecha-nism can provide the innovator with some quasi-monopoly protection from imitators.

No formal studies yet exist, but the value of response time to innovators can bereasoned to be a function of various situation-specific factors. For example, considerthe effect of the length of response time divided by length of the customer purchasedecision cycle. A high value of this factor favors the innovator over imitators. Considerone extreme example: a consumer fad item (very short purchase decision time) thatsells in high volume for six months only. Assume that the item can be readily imitated

73


but can only be produced economically by mass-production tooling that requires sixmonths to build. Obviously, response time here allows the innovator to monopolizethe entire market if he can supply it with his initial tooling. At another extreme is anexpensive capital-equipment innovation that customers typically take two years todecide to buy, budget for, and so on-and that competitors can imitate in one year.Obviously, response time in this instance affords an innovator little protection.

2. The source of figures in this section is William G. Lionetta, Jr., "Sources ofInnovation Within the Pultrusion Industry" (SM thesis, Sloan School of Management,MIT, Cambridge, Mass., 1977), chap. 2.

3. A homebuilt machine of average capacity (i.e., a machine capable of pultrudingproduct with a cross-section of 6 in. by 7 in.) had a direct cost of at most $60,000 in1977 (Lionetta, "Sources of Innovation Within the Pultrusion Industry," 43) versus theprice of $95,000 charged for an equivalent commercial machine. Presumably the user-built machines were cheaper because the user does not incur some expenses that themachine builder must, such as selling expenses.

4. C. F. Pratten, Economies of Scale in Manufacturing Industry (Cambridge: Cam-bridge University Press, 1971), Table 17-4, p. 173.

5. The primary exception was four-wheel drive. Although it offered advantages toall users, it was a costly feature most advantageous to those operating on difficultterrain and, so, penetrated the market more slowly. The inclusion of four-wheel drivein a tractor shovel added over $1000 in direct cost per unit at the time of the innova-tion.

6. This estimate of operating savings is derived in the following manner. Standardindustry assumptions are that the life of a tractor shovel is five years and that it willoperate 2000 hr/yr. Productivity savings primarily involve savings of labor and capital.If we assume an operator was paid $1/hr plus 50% fringe and overhead, which wasaverage at the time of the innovations (Council of Economic Advisers Annual Report1981, in Economic Report of the President, Transmitted to the Congress, January 1981,Together with the Annual Report of the Council of Economic Advisers [Washington,D.C.: U.S. Government Printing Office, 1981], 21-213), an innovation that made atractor shovel 20% more productive-a reasonable estimate for the major improve-ment innovations studied-would involve a savings of 20% of an operator's salary (or$600/yr) and a savings of 20% the capital cost of a machine. Since, at the time of theinnovations, tractor shovel prices ranged from $5000 to $10,000/unit (or a maximum of$2000 depreciated over five years straight line), total maximum annual savings to theuser in capital and labor therefore were $1100/yr/machine/innovation.

7. Glen L. Urban et al., "Market Share Rewards to Pioneering Brands: An Empiri-cal Analysis and Strategic Implications," Management Science 32, no. 6 (June 1986):645-59.

8. Pratten, Economies of Scale in Manufacturing Industry, Table 17-4, p. 173.9. C. T. Taylor and Z. A. Silberston, The Economic Impact of the Patent System:

A Study of the British Experience (Cambridge: Cambridge University Press, 1973).10. James A. Rauch, ed., The Kline Guide to the Plastics Industry (Fairfield, N.J.:

Charles H. Kline, 1978), Table 3-4, p. 59.11. Herbert Solow, "Delrin: Du Pont's Challenge to Metals," Fortune 60, no. 2

(August 1959): 116-19.12. Ibid.13. Marshall Sittig, PolyAcetyl Resins (Houston, Tex.: Gulf Publishing Company,

1963).14. National Academy of Sciences, Applied Science and Technological Progress, A

74


report to the Committee on Science and Astronautics, U.S. House of Representatives,GP-67-0399 (Washington, D.C.: U.S. Government Printing Office, June 1967), 37.

15. Wickam Skinner and David C. D. Rogers, Manufacturing Policy in the Electron-ics Industry: A Casebook of Major Production Problems, 3rd ed. (Homewood, Ill.:Richard D. Irwin, 1968).

16. Pieter A. VanderWerf, "Explaining the Occurrence of Industrial Process Inno-vation by Materials Suppliers with the Economic Benefits Realizable from Innovation"(Ph.D. diss., Sloan School of Management, MIT, Cambridge, Mass., 1984).

17. Robert K. Merton, The Sociology of Science: Theoretical and Empirical Investi-gations, ed. Norman W. Storer (Chicago: University of Chicago Press, 1973).

18. Glen L. Urban and John R. Hauser, Design and Marketing of New Products(Englewood Cliffs, N.J.: Prentice-Hall, 1980).

19. Success at this task will allow us to lessen the amount of detail we need andexploit the possibilities of larger data bases such as those recently developed by KeithPavitt and F. M. Scherer. Pavitt recently examined a data base of over 2000 Britishinnovations to begin to map the origins and flow of technology between sectors ofBritish industry ("Sectoral Patterns of Technical Change: Towards a Taxonomy and aTheory," Research Policy 13, no. 6 [December 1984]: 343-73). Scherer recently gener-ated and studied a data base on the interindustry flows of technology based on datafrom 443 large U.S. industrial companies (Innovation and Growth: SchumpeterianPerspectives [Cambridge, Mass.: MIT Press, 1984], 51).

20. George J. Stigler, The Organization of Industry (Homewood, Ill.: Richard D.Irwin, 1968).

21. Robert H. Hayes and Steven C. Wheelwright, Restoring Our Competitive Edge:Competing Through Manufacturing (New York: Wiley, 1984), 381.

testing the relationship between the functional source of ... · the functional source of...

Documents